NCAR's Research Data Archive: OPeNDAP Access for Complex Datasets

Abstract:

Many datasets have complex structures including hundreds of parameters and numerous vertical levels, grid resolutions, and temporal products. Making these data accessible is a challenge for a data provider. OPeNDAP is powerful protocol for delivering in real-time multi-file datasets that can be ingested by many analysis and visualization tools, but for these datasets there are too many choices about how to aggregate. Simple aggregation schemes can fail to support, or at least make it very challenging, for many potential studies based on complex datasets. We address this issue by using a rich file content metadata collection to create a real-time customized OPeNDAP service to match the full suite of access possibilities for complex datasets.

The Climate Forecast System Reanalysis (CFSR) and it’s extension, the Climate Forecast System Version 2 (CFSv2) datasets produced by the National Centers for Environmental Prediction (NCEP) and hosted by the Research Data Archive (RDA) at the Computational and Information Systems Laboratory (CISL) at NCAR are examples of complex datasets that are difficult to aggregate with existing data server software. CFSR and CFSv2 contain 141 distinct parameters on 152 vertical levels, six grid resolutions and 36 products (analyses, n-hour forecasts, multi-hour averages, etc.) where not all parameter/level combinations are available at all grid resolution/product combinations. These data are archived in the RDA with the data structure provided by the producer; no additional re-organization or aggregation have been applied. Since 2011, users have been able to request customized subsets (e.g. - temporal, parameter, spatial) from the CFSR/CFSv2, which are processed in delayed-mode and then downloaded to a user’s system. Until now, the complexity has made it difficult to provide real-time OPeNDAP access to the data. We have developed a service that leverages the already-existing subsetting interface and allows users to create a virtual dataset with its own structure (das, dds). The user receives a URL to the customized dataset that can be used by existing tools to ingest, analyze, and visualize the data. This presentation will detail the metadata system and OPeNDAP server that enable user-customized real-time access and show an example of how a visualization tool can access the data.